基于多分辨率的快速迭代最近点配准算法

针对迭代最近点(Iterative Closest Point,ICP)算法计算时间长的问题,提出一种基于多分辨率配准点的ICP算法。使用自适应体素网格滤波器对原始点云进行多分辨率采样,利用低分辨率点云快速迭代获得两点云间初始变换矩阵;利用高分辨率点云在初始变换基础上做更精确配准。实验结果表明,该算法在配准精度基本不变的情况下,可以显著降低配准时间,且随着点云点数增加,速度提升效果越明显。

FAST ITERATIVE CLOSEST POINT REGISTRATION ALGORITHM BASED ON MULTI-RESOLUTION

Aiming at the problem of long computation time of the iterative closest point(ICP)algorithm,we propose an ICP algorithm based on multi-resolution registration.It used the adaptive voxel grid filter to perform multi-resolution sampling of the original point cloud.The initial transformation matrix between two clouds was obtained by fast iteration of low-resolution point clouds.Then,the high-resolution point clouds were used for more accurate registration based on the initial transformation.The experimental results show that the algorithm can significantly reduce the registration time when the accuracy is basically unchanged.And as the number of point cloud points increases,the speed improvement effect becomes more obvious.